New framework examines AI systems as hidden cognitive infrastructure reshaping human thought

AI systems increasingly function as invisible foundations that structure what people can know and do in the digital world, yet mainstream human-computer interaction research has largely overlooked this transformative role. A new research framework seeks to remedy this gap by treating AI as "cognitive infrastructure"—foundational systems that operate beneath conscious awareness.

Reframing AI as Hidden Infrastructure#

The concept of Cognitive Infrastructure Studies (CIS) proposes a shift in how we analyze AI's impact on human cognition. Rather than focusing narrowly on user interfaces or interaction design, this approach treats AI systems as semantic infrastructures that transport meaning and shape thought at the population scale. These systems work through what researchers call anticipatory personalization and adaptive invisibility—they learn what feels relevant to each user while remaining so seamlessly integrated into daily digital life that their influence becomes nearly impossible to detect.

A central concern is how AI automates "relevance judgment," effectively transferring authority over what information matters from humans to algorithmic systems. This shift moves the "locus of epistemic agency"—the power to determine what counts as knowledge—away from people and toward machines.

Three Layers of Cognitive Impact#

The framework traces AI's influence across interconnected scales:

• Individual level: Users develop cognitive dependency as they habitually rely on AI preprocessing to filter, interpret, and present information
• Collective level: Algorithmic shaping of information environments affects democratic deliberation and public reasoning
• Societal level: AI preprocessing influences broader governance structures and cultural epistemologies—the shared ways societies determine what is true or knowable

Bridging Disciplinary Gaps#

A key innovation of CIS is its attempt to connect insights across fields that typically operate in isolation. Cognitive science can analyze individual mental processes but lacks tools for studying preprocessing effects at population scale. Digital sociology examines societal patterns but cannot access the individual cognitive mechanisms underlying them. Meanwhile, computational approaches often miss how algorithmic influence propagates through cultural transmission.